This proposal brings together new data from Montreal and new theory from Michigan to generate a new understanding of how HIV spreads, of how spreading patterns alter the effects of control programs, and of how genetic analyses can be used to predict control results in any particular population. The Montreal data shows a high rate of clustering in HIV genetic sequences from early infection but little clustering from late infection. The clusters represent active transmission chains that might be detected and interrupted. The Michigan theory shows that risk behavior fluctuations interact strongly with high PHI transmission to affect transmission dynamics, the pattern of clustering, and the effects of HIV transmission control programs. The pattern of HIV clustering should therefore help predict transmission control program effects. We will define effective HIV control programs that balance efforts to detect and interrupt active transmission foci with efforts to decrease transmission through earlier treatment. That balance will differ when transmission is diffuse and constant or when it is punctuated by mini-epidemics of transmission during primary infection. The high rate of early diagnosis and virus sequence analysis in Montreal gives us the most refined view to date of the spatial-temporal dynamics of infection. The model analyses we now propose will greatly refine that view and clarify what effects different control strategies will have under different conditions. Our preliminary results show that behavioral risk fluctuations change expected intervention effects from both therapy and from interrupting active transmission chains. Fortunately, these effects are reflected by genetic sequence patterns. Accordingly we pursue 3 Aims. Aim 1 maps control program effects across a broad range of contact patterns, population structures, risk behaviors, and natural history of infection variants. Aim 2 maps transmission tree patterns on to that same space. This will allow interpretation of genetic patterns from anywhere where such patterns might be gathered. Aim 3 adapts models specifically to Montreal by fitting model parameter values in a two stage estimation process. The first stage fits deterministic compartmental models to HIV surveillance and special study data using an MCMC process. The second stage refines the parameter space estimated in the first stage by fitting genetic sequences on to transmission trees generated by a stochastic process model that has been shown on average to exactly reproduce the deterministic model results. In this refined parameter space the methods of Aim 2 indicate the consequences of various alternatives in HIV control program design. PUBLIC HEALTH RELEVANCE: Powerful new data are uniquely available from Montreal on how HIV transmissions during early infection are genetically, spatially, and temporally clustered. These will be analyzed using new theory developed at Michigan to determine how infection is flowing through a population and how early HIV Rx and traditional Public Health STD control actions can stop that flow. The new insights and analysis methods developed should help all control programs everywhere set more effective control priorities.